Infected V. borleae, V. davyi, V. exuvialis, V. hebeclada, V. natalitia, V. permixta, V. swazica and V. xanthophloea trees were sampled during several field surveys in South Africa between 2004 and 2015. Leaves bearing uredinial and telial rust sori and short branches having aecial galls were collected and subsequently dried between paper sheets in a plant press. In total, 49 specimens were studied based on morphology and 31 of these could be used for phylogenetic analyses based on DNA-sequence data. Ten of the 49 specimens were either type or voucher specimens collected in the late 19^th and early 20^th century and used by Doidge for her studies of the southern African Ravenelia spp. (Doidge 1927, 1939). These herbarium specimens were used only for morphological comparisons with the newly collected material (Table 1). Three specimens were deposited in KR, all others in PREM.

Spores from individual sori were collected separately using sterile insect needles. Genomic DNA extractions were made using the INNUPrep Plant DNA Kit (Analytik Jena, Germany) following the manufacturer’s protocols with the following modifications: Spores were crushed using a Retsch mixer mill MM2000 (Retsch, Haan, Germany) by shaking them together with 2 steel beads of 2.5 mm diameter in a 2.0 ml Eppendorf tube. This process was repeated in three consecutive cycles. In the first step, the closed tubes were cooled in liquid nitrogen and immediately shaken for 2 min at 100 Hz. Thereafter 10–40 μl of lysis buffer was added to the tube to loosen spore remnants from the inner side of the Eppendorf tube lid using a vortex mixer followed by a centrifugation step. Samples were again cooled in liquid nitrogen and shaken for an additional 2 min at 100 Hz followed by centrifugation for 1 min at 6000 rcf. The last two steps were repeated once.

For PCR of the ribosomal nrITS and LSU rDNA gene regions, the Taq-DNA-Polymerase Mix (PeqLab, Erlangen, Germany) was used with the primers ITS1F (Gardes and Bruns 1993) and RustITS1F (Toome and Aime 2014), respectively and ITS4BR (Vialle et al. 2009). PCR of the LSU rDNA region was performed using the primer pairs LR0R and LR6 (Vilgalys and Heester 1990). Two additional primers (5.8SrustF: 5’- CGA TGA AGA ACA CAG TGA AAT GTG; D1D2RustR: 5’- CTY TGC TAT CCT GAG GGA) were designed with improved specificity for Ravenelia and that reduced amplification of ascomycetous non-target organisms. The thermal cycling conditions for primers ITS1F/ITS4BR and RustITS1F/ITS4BR were as follows: 2 min at 96 °C followed by 40 cycles of 20 sec at 96 °C, 40 sec at 50 °C and 50 sec at 72 °C, final extension was for 5 min at 72 °C; for primers LR0R/LR6: 3 min at 96 °C followed by 40 cycles of 30 sec at 95 °C, 40 sec at 49 °C and 1 min at 72 °C, final extension was for 7 min at 72 °C; for primers 5.8SrustF/D1D2rustR: 3 min at 96 °C followed by 40 cycles of 30 sec at 96 °C, 45 sec at 54 °C and 1 min 20 sec at 72 °C, final extension was for 7 min at 72 °C.

PCR products were purified using Sephadex G-50 columns (Sigma-Aldrich, Steinheim, Germany). Where PCR products showed only weak bands on agarose gels, purification was undertaken using the Zymo Research DNA Clean & Concentrator^TM-5 Kit (Zymo Research GmbH, Freiburg, Germany) following the manufacturer’s protocol. DNA sequencing was carried out in both directions using the same primers as those used for PCR on a 3130XL Genetic Analyzer (Applied Biosystems) at the sequencing service of the Faculty of Chemistry and Biochemistry of the Ruhr University Bochum, Germany.

Sequences were screened against the NCBI GenBank using the BLASTn algorithm (Altschul et al. 1990) to exclude erroneously amplified contaminants from further processing. Forward and reverse strands of the rust sequences were assembled using Sequencher 5.0 software (Gene Codes Corp., Ann Arbor, MI, USA) and, where necessary, manually edited. In total, 32 sequences were used to construct an alignment of the nrITS and LSU rDNA sequence data, respectively, using MAFFT v6.832b (Katoh and Standley 2014) applying the L-INS-i strategy. Maximum likelihood analyses were conducted in RAxMLGUI v.1.3 (Silvestro and Michalak 2012) using RAxML 8.0.26 (Stamatakis 2014) using the general time reversible model of nucleotide substitution (Lanave et al. 1984) with gamma distributed substitution rates. The analyses were run with a rapid bootstrap analysis using 1000 bootstrap replicates. The analyses were first conducted for each dataset separately and topological congruence was checked visually. As no conflict of supported phylogenetic groupings was observed, the final phylogeny was inferred by combining both datasets of the nrITS and LSU rDNA sequences applying the same methodology as for individual datasets.

Parsimony network analyses were performed using TCS v1.21 (Clement et al. 2000) and the same sequence alignments that were used for the phylogenetic analyses. Gaps were deleted from calculations and the default connection limit of 95% was used.

The spores of the dried herbarium specimens (Table 1) were scraped from leaf surfaces and mounted in lactophenol on microscope slides. A minimum of ten teliospores and 30 urediniospores per specimen were examined. Minimum, maximum and mean values are provided in Table 2. The specimens PREM2211 (R. evansii), PREM2403 (R. evansii), PREM2539 (R. evansii), PREM6807 (R. evansii), PREM7105 (R. evansii), PREM1935 (R. natalensis), PREM2514 (R. natalensis), PREM2375 (R. glabra), PREM10698 (R. glabra) and PREM20727 (R. glabra) were examined at the facilities of the ARC-Plant Protection Institute (ARC-PPRI), Roodeplaat, South Africa using a Leica Dialux 22 EB microscope and a ColorView III CCD colour camera. Measurements were made using analySIS LS software (LS Research Software GmbH, Germany). The remaining specimens were studied at the Ruhr University Bochum, Germany, using a Zeiss Axioplan light microscope. Morphological characteristics were measured using Cell^D v. 3.1 imaging software (Olympus Soft Imaging Solutions GmbH, Germany) and Zen2 lite (Blue Edition) V. 2.0.0.0 (Carl Zeiss Microscopy, 2011, Jena, Germany). Photographs were obtained using a Color View microscope camera (Olympus Soft Imaging System, Germany). For detailed investigations of the spore-surface structures, scanning electron microscopy (SEM) was used. For this purpose, infected leaflets from the herbarium specimens were mounted on double-sided adhesive carbon tape on metal stubs and coated with gold in a sputter coater BAL-TEC SCD OSO (Capovani Brothers Inc, USA). Subsequently, the samples were examined using a ZEISS Sigma VP scanning electron microscope.

For principal component analyses (PCA) the morphological data collected for all examined rust individuals were separated into sub-sets based on preliminary species assignments representing R. evansii, R. macowaniana and Ravenelia sp. (Groups A, B1 and B2). PCA for all subsets was conducted separately using the R-packages plyr and ggplot2 implemented in R (www.R-Project.org). Six characteristics of teliospores providing numeric data were defined and measured: teliospore diameter, probasidial cell length and probasidial cell width, number of cells in diameter, epispore thickness and ornamentation length. Mean values were calculated for the individual teliospore measurements, scaled and missing values were deleted from analyses.

Sequence data from the nrITS and LSU rDNA gene regions were obtained for all 31 newly collected specimens. The alignment of the nrITS sequence dataset had a total length of 764 bp with 133 variable sites of which 131 positions were parsimony informative. The aligned sequences of the LSU rDNA dataset had a length of 922 bases and comprised 31 sequences with 45 variable sites and 40 parsimony informative positions. The combination of the nrITS and LSU rDNA datasets resulted in an alignment with a total length of 1686 nucleotides comprising 32 sequences. The sequence alignment and phylogenetic tree of the combined rDNA sequence data set was deposited at TreeBASE (http://purl.org/phylo; submission IDS22307).

Maximum likelihood analysis of the combined dataset resulted in a phylogenetic tree that consisted of three highly supported groups representing R. evansii, R. macowaniana and a novel Ravenelia species described below (Fig. 1A).

Figure 1. 

Phylogenetic reconstruction of Ravenelia species on different Vachellia hosts A Maximum likelihood tree with 1000 bootstrap repeats based on combined nrITS and LSU rDNA sequence data. Bootstrap values below 75 are not shown. Three highly supported groups represent R. evansii, R. macowaniana and R. xanthophloeae sp. nov., respectively. Specimens that originated from formerly unreported host species are highlighted in bold B Parsimony network analysis based on the same dataset as in the ML-analysis. Each line represents one base substitution while small circles represent intermediate but missing sequences. Numbers next to lines indicate the positions of the substitutions in the alignment. Sequences in rectangular boxes were inferred as ancestral by this analysis.

The phylogenetic group representing R. evansii consisted of 17 sequences. Three of these were obtained from Vachellia sieberiana var. woodii (PREM61223, PREM61228, PREM61881) and three from V. robusta ssp. robusta (KR-M-43649, PREM61208, PREM61209). These are tree species that had previously been reported as hosts of R. evansii. Rust specimens collected from the following five Vachellia species also clustered in this group: V. borleae (ME384 and PREM61869), V. davyi (PREM61005), V. exuvialis (PREM61868, PREM61876), V. hebeclada (PREM61227) and V. swazica (PREM61002, PREM61008, PREM61028). These are all newly reported hosts for R. evansii.

A second group included the sequences of eleven specimens, five of which originated from V. karroo and were identified as R. macowaniana (KR-M-43657, PREM61222, PREM61221, PREM61210, KR-M-43406). Five specimens were collected from V. natalitia (PREM61214, PREM61862, PREM61218, PREM61219, PREM61216) and one originating from V. permixta (PREM61875) also clustered in this group. The latter two hosts are newly reported for R. macowaniana.

A distinct clade, nested within the R. macowaniana group, was represented by three Ravenelia specimens that were isolated from V. xanthophloea (PREM61215, PREM61213, PREM61000) suggesting that it represents a novel taxon. For PREM61213, two identical sequences were obtained, one derived from aeciospores and one from teliospores.

The parsimony network analysis, based on the combined set of nrITS and LSU rDNA sequence data, separated three distinct groups each comprising the same specimens representing R. evansii, R. macowaniana and the novel Ravenelia species in our phylogenetic analysis, respectively (Fig. 1B). Network analysis relying on LSU alone could not separate R. macowaniana from the novel taxon, while separation of these two groups was observed based on nrITS alone (not shown). The R. evansii group consisted of six haplotypes of 17 sequences that differed by a maximum of two substitutions from the inferred ancestral sequences (PREM61005 and PREM61227). In R. macowaniana, sequence divergence was more pronounced and consisted of nine haplotypes in a total of eleven sequences. The highest rate of six substitutions was observed for specimen PREM61221 relative to the inferred ancestral sequence (KR-M-43657). Specimens collected on V. xanthophloea had only one substitution.

Ravenelia evansii

The teliospore morphology of R. evansii specimens showed a considerable overall variability in all six investigated teliospore characteristics (Suppl. material 1: Fig. S1, Table 2). The voucher specimens PREM61869 (on V. borleae), PREM61876 and PREM61868 (both on V. exuvialis) had significantly smaller teliospores compared to the remaining specimens, but variation in this trait could also be observed within single host associations, e.g. within those from V. davyi and V. robusta (Suppl. material 1: Fig. S1).

The principal component analysis (PCA) of teliospore characteristics clustered several individuals derived from specific hosts into distinct groups (Fig. 2C, D). Individuals that originated from V. borleae and V. exuvialis clustered more closely and could be separated from those individuals that were collected from V. davyi (Fig. 2D). The separation of these individuals was supported by PC1 which could explain 37.6% of the overall variability. The traits ‘cells in diameter’ and ‘teliospore diameter’ were characteristics that corresponded best with this axis (Fig. 2C). These results indicated, that the latter two characters were the most variable traits to separate these individuals with different host associations. However, individuals from V. luederitzii var. retinens, V. robusta var. robusta, V. sieberiana var. woodii and V. swazica showed only weak separation, i.e. less clear patterns of morphological separation.

Figure 2. 

Biplots of a principal component analysis (PCA) of six teliospore characteristics of specimens of A Ravenelia macowaniana originating from Vachellia karroo (red) and V. natalitia (green) and B in comparison with R. xanthophloeae sp. nov. collected from V. xanthophloea (blue) C, D represent R. evansii originating from seven distinct Vachellia species. Each dot represents an individual teliospore for which mean values of multiple measurements of all six defined morphological characteristics were calculated. Each colour represents the host species of the individual rust specimen. In D only spore representatives collected from V. borleae, V. exuvialis and V. davyi were highlighted to gain better visibility.

Ravenelia macowaniana

Specimens representing R. macowaniana were morphologically more homogeneous compared to R. evansii. Here, only spore characteristics such as ‘probasidial cell width’ and ‘epispore thickness’ were often significantly different between investigated specimens (Suppl. material 2: Fig. S2). There was little variation for the characters ‘teliospore diameter’, ‘probasidial cell length’ and the number of ‘cells in diameter’. These characters varied distinctly only between single specimens, e.g. PREM61226 originating from V. natalitia had significantly larger teliospores in comparison to specimen PREM61888 that was also collected from this host (Suppl. material 2: Fig. S2).

For the specimens of R. macowaniana, PC1 and PC2 could explain 36.2% and 22.4% of the similarity, respectively. However, unlike in R. evansii, single teliospore characteristics did not differ significantly in terms of host association (Fig. 2A).

Ravenelia sp. nov.

Due to similar teliospore characteristics, R. macowaniana was compared using PCA to individuals of the undescribed Ravenelia species collected on V. xanthophloea in order to characterise and, if possible, to contrast both morphologies. The PCA separated two groups that corresponded well with R. macowaniana and the novel Ravenelia species and showed very little overlap in morphological characteristics (Fig. 2B). Separation between both species groups was mostly seen in PC1 that could explain 42.5% of the overall variability with the ‘ornamentation length’ corresponding best to this axis (Fig. 2B). Although less distinct in the PCA, mean values of teliospore characters also revealed that the characters ‘epispore thickness’, ‘probasidial cell length’ and especially the ‘teliospore diameter’ and ‘cells in diameter’ are valuable characters for the discrimination of both species (Fig. 3). All spore measurements, derived from the six defined spore characteristics that were used for PCA, are available as an excel-file in the Suppl. materials 3, 4: tables S1, S2: ‘PCA-table R. evansii’ and ‘PCA-table R. macowaniana’, respectively.

Figure 3. 

Radarchart of mean values of the morphological investigations of teliospore characteristics of Ravenelia macowaniana originated from Vachellia karroo (red), V. natalitia (green) and R. xanthophloeae on V. xanthophloea (blue). Numbers on y-axis represent the respective minimum and maximum values. This radarchart reveals the morphological differences between R. macowaniana and R. xanthophloeae.